Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.88%. Comparing base (a4b8656) to head (e0cdb73).
⚠️ Report is 1 commits behind head on main.

@@            Coverage Diff             @@
##             main     #809      +/-   ##
==========================================
+ Coverage   97.87%   97.88%   +0.01%     
==========================================
  Files         639      643       +4     
  Lines       70008    70418     +410     
==========================================
+ Hits        68517    68926     +409     
- Misses       1491     1492       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

Agentic Review Report

Structural Check

Structural Review: model+rule SchedulingToMinimizeWeightedCompletionTime

Structural Completeness

Build Status

• make test: PASS
• make clippy: PASS

Semantic Review

• Model evaluate(): OK — compute_weighted_completion_time() groups tasks by processor and applies Smith's rule via exact cross-product comparison, then accumulates the weighted completion objective with overflow checks (src/models/misc/scheduling_to_minimize_weighted_completion_time.rs:145).
• Model dims(): OK — vec![self.num_processors; self.num_tasks()] matches the issue's assignment-only configuration space (src/models/misc/scheduling_to_minimize_weighted_completion_time.rs:215).
• Size getters / overhead names: OK — num_tasks() and num_processors() exist and match the complexity / overhead symbols used by the model and rule (src/models/misc/scheduling_to_minimize_weighted_completion_time.rs:111, src/models/misc/scheduling_to_minimize_weighted_completion_time.rs:117).
• Weight handling: OK — the model keeps weights as inherent fields/getters rather than introducing extra traits (src/models/misc/scheduling_to_minimize_weighted_completion_time.rs:126).
• extract_solution(): OK for valid ILP outputs — it decodes the one-hot assignment block back to per-task processor indices (src/rules/schedulingtominimizeweightedcompletiontime_ilp.rs:54).
• Overhead accuracy: ISSUE — the implemented ILP omits the C_t <= M bounds required by issue #783 step 6. The code adds only C_t >= l_t at src/rules/schedulingtominimizeweightedcompletiontime_ilp.rs:107, and the emitted canonical bundle has 25 variables / 70 constraints rather than the 75 documented in the issue packet. The current overhead formula at src/rules/schedulingtominimizeweightedcompletiontime_ilp.rs:68 matches the implementation, but not the issue specification.
• Paper quality: ISSUE — the main model definition introduces the assignment map p : T -> {1, ..., m} and then immediately defines C(t) = sigma(t) + ell(t) without defining sigma, so the mathematical statement is internally inconsistent as written (docs/paper/reductions.typ:5595).

Issue Compliance

Summary

• 28/28 structural checklist items passed.
• make test clippy passed in the review worktree.
• ISSUE: src/rules/schedulingtominimizeweightedcompletiontime_ilp.rs:107 omits the C_t <= M bounds promised by issue #783, which changes the documented constraint count from 75 to 70 on the canonical example.
• ISSUE: docs/paper/reductions.typ:5595 uses sigma(t) without defining it after introducing p as the assignment map.

Quality Check

Quality Review

Design Principles

• DRY: OK — the model/rule follow existing scheduling and ILP patterns without obvious copy-paste debt beyond project-standard create/help boilerplate.
• KISS: OK — the model stays assignment-only, and the rule uses a straightforward x/C/y layout without unnecessary abstraction.
• HC/LC: OK — model logic, rule logic, tests, CLI registration, and paper/docs remain separated cleanly by module.

HCI (CLI changed)

• Error messages: OK — e.g. missing --weights reports an actionable usage string for the new create path.
• Discoverability: ISSUE — pred reduce --help still advertises pred reduce problem.json --to ILP -o reduced.json (problemreductions-cli/src/cli.rs:790), but this feature only works when the target is spelled ILP/i32.
• Consistency: ISSUE — pred solve /tmp/scheduling_wct.json auto-reduces successfully to ILP, while the explicit pred reduce ... --to ILP and pred path ... ILP flows fail for the same problem unless the user adds /i32.
• Least surprise: ISSUE — the fallback hint in problemreductions-cli/src/commands/reduce.rs:129 repeats the non-working plain-ILP recipe (pred path <SRC> ILP -o path.json), so a user who follows the hint hits another failure immediately.
• Feedback: OK — successful create/solve/reduce commands report outputs clearly and the explicit ILP/i32 flow works end to end.

Test Quality

• Naive test detection: ISSUE
  • src/unit_tests/rules/schedulingtominimizeweightedcompletiontime_ilp.rs:9 checks num_vars and objective coefficients, but nothing asserts the documented completion upper bounds or the expected 75-constraint canonical example. That gap is why the C_t <= M drift is currently unguarded.

Issues

Critical (Must Fix)

• None.

Important (Should Fix)

• src/rules/schedulingtominimizeweightedcompletiontime_ilp.rs:107 / src/rules/schedulingtominimizeweightedcompletiontime_ilp.rs:68: the ILP rule omits the C_t <= M bounds from issue #783, so the implemented reduction no longer matches the documented size formula/example count.
• problemreductions-cli/src/cli.rs:790 / problemreductions-cli/src/commands/reduce.rs:129: the explicit reduce/path UX tells users to target plain ILP, but the new rule is only reachable through ILP/i32; the generated hint currently loops the user into another failing command.
• src/unit_tests/rules/schedulingtominimizeweightedcompletiontime_ilp.rs:9: the rule tests do not pin the missing-bound / constraint-count behavior, so the spec drift above is not caught automatically.

Minor (Nice to Have)

• docs/paper/reductions.typ:5595: rename or define the scheduling/start-time function consistently instead of mixing p and sigma in the same sentence.

Summary

• The implementation is structurally clean and all build/test gates passed.
• The main correctness/spec issue is the missing C_t <= M family in the ILP reduction relative to issue #783.
• The main UX issue is that explicit reduce/path guidance for ILP targets is still variant-insensitive, even though this feature needs ILP/i32.

Agentic Feature Tests

Feature Test Report: problemreductions

Date: 2026-03-29
Project type: CLI + library
Features tested: SchedulingToMinimizeWeightedCompletionTime model discovery/create/solve and explicit -> ILP/i32 reduction
Profile: ephemeral
Use Case: downstream CLI user wants to discover the new scheduling model from the catalog, create the canonical example, solve it, then explicitly reduce it to ILP and solve the reduction bundle.
Expected Outcome: the model should be discoverable from CLI docs/listing, the example should build, direct solve should return the expected optimum 47, and the explicit ILP reduction path should be runnable from docs/help without guesswork.
Verdict: fail
Critical Issues: 0

Summary

Per-Feature Details

SchedulingToMinimizeWeightedCompletionTime model + ILP rule

• Role: downstream CLI user reading built-in help and catalog output.
• Use Case: inspect the new model, create the canonical example, solve it directly, then reduce it to ILP and solve the bundle.
• What they tried:
  • ./target/debug/pred list
  • ./target/debug/pred show SchedulingToMinimizeWeightedCompletionTime
  • ./target/debug/pred create --example SchedulingToMinimizeWeightedCompletionTime -o /tmp/scheduling_wct.json
  • ./target/debug/pred solve /tmp/scheduling_wct.json
  • ./target/debug/pred reduce /tmp/scheduling_wct.json --to ILP
  • ./target/debug/pred path SchedulingToMinimizeWeightedCompletionTime ILP
  • ./target/debug/pred path SchedulingToMinimizeWeightedCompletionTime ILP/i32
  • ./target/debug/pred reduce /tmp/scheduling_wct.json --to ILP/i32 -o /tmp/scheduling_wct_ilp.json
  • ./target/debug/pred solve /tmp/scheduling_wct_ilp.json --solver ilp
• Discoverability: Partial — the model appears correctly in pred list, pred show is informative, and the outgoing reduction is visible; the friction starts when the user tries to run the explicit ILP reduction from generic help text.
• Setup: Yes — the built workspace binary was sufficient; no external setup beyond the repo build was needed.
• Functionality: Yes — direct solve returned Min(47) with solution [0,1,0,1,0], and the explicit ILP/i32 reduction bundle also solved back to the same optimum.
• Expected vs Actual Outcome: Partial — the feature works end to end, but only after the user guesses ILP/i32; the advertised plain ILP spelling fails for explicit reduce/path flows.
• Blocked steps: None.
• Friction points:
  • pred reduce /tmp/scheduling_wct.json --to ILP fails with No witness-capable reduction path ... even though the direct rule exists.
  • The failure hint recommends pred path SchedulingToMinimizeWeightedCompletionTime ILP -o path.json, which also fails; the working target spelling is ILP/i32.
• Doc suggestions: preserve target variants in explicit reduce/path help and error hints, or suggest pred show ILP/i32 / pred path <SRC> ILP/i32 when the target has no variant-free witness path.

Expected vs Actual Outcome

• Expected: a user could follow catalog/help output to explicitly reduce the new scheduling model to ILP.
• Actual: the model is discoverable and fully functional, but the explicit reduction path is only discoverable if the user already knows to use ILP/i32 instead of the plain ILP spelling shown in help/error text.

Issues Found

• Important (confirmed): explicit pred reduce/path ... ILP guidance is misleading in the current worktree; the working target is ILP/i32.
• Minor (confirmed): inspecting the generated canonical reduction bundle shows 25 variables / 70 constraints, not the 75-constraint example stated in issue #783 and the review packet.

Suggestions

• Update explicit reduce/path help and hints to preserve target variants.
• Add a rule-level assertion for the canonical constraint count / completion upper bounds so the ILP formulation matches its issue specification.

Generated by review-pipeline